Link: Machine learning

What is bias variance trade off

Beginner’s error: overfit model

  • Beginner’s error: make model more complicated and “overfit”
  • As a result, the model is overfit and performs poorly when forecasting with the test dataset

How to evaluate?

Plot the complicity/flexibility of the model, by the polynomial level of a regression fit vs error metrics (Mean Squared Error MSE).

Find the balanced point

Find the point that’s good enough to balance out the bias and the variance